RAI Evaluation

AI Incidents Rising and Responsible AI Evals Still Rare Among Major Developers, Stanford HAI 2025 Index Finds

Stanford University's Human-Centered Artificial Intelligence institute published its 2025 AI Index Report on April 1, 2025, providing a global analysis of AI research, development, and governance trends. The report documents an increase in AI-related incidents and finds that standardized responsible AI evaluations remain rare among major industrial model developers, identifying a gap between organizational recognition of RAI risks and concrete action. New safety and factuality benchmarks including HELM Safety, AIR-Bench, and FACTS are highlighted as emerging tools for assessing model behavior, though adoption is limited. Governments across multiple jurisdictions accelerated regulatory output during the period covered, with frameworks from the OECD, EU, and United Nations emphasizing transparency and trustworthiness requirements. For enterprise compliance teams, the report reinforces pressure to formalize RAI evaluation processes and signals that regulators are moving from principle-setting toward enforceable standards. Organizations that have not yet aligned internal AI governance practices with emerging benchmarks and government frameworks face increasing exposure as scrutiny from regulators and auditors intensifies.